NOVEL MORPHOLOGICAL ANALYSIS OF A FOSSIL FABACEAE POLLEN TYPE, STRIATOPOLLIS CATATUMBUS (TRIBE DETARIAE)

Striatopollis catatumbus is a pantropical fossil pollen grain known from the early Paleocene to the Holocene (65-0.01 Ma) and associated with fluvial-deltaic depositional environments. Previous studies have suggested that the morphotype may be related to the Fabaceae tribe Detariae, which includes the genera Macrolobium, Anthonotha, Isoberlinia, and Crudia. In this study, we use two novel methods to compare the morphology of S. catatumbus to 15 Detariae genera with similarly striate pollen. First, we use Airyscan confocal superresolution microscopy, which allows us to visualize both the external and internal 3D morphology of the pollen wall at a resolution comparable to scanning and transmission electron microscopy. Second, we introduce the use of deep-learning computer vision algorithms to virtually reconstruct ancestral morphologies based on visual shape and texture features extracted from images of known samples. These algorithms produce explicit hypotheses of ancestral morphology and provide a reproducible, computational method for placing our fossil specimens on the Detariae phylogeny.

We identify three different fossil morphotypes of striate pollen that are currently classified as S. catatumbus, two in the Neotropics and one in Africa. Our results suggest a relationship between the phylogenetic radiation of the modern genera of Detariae and the biogeographic expansion of S. catatumbus across the tropics during the Eocene. We also demonstrate that modern Detariae genera are restricted to tropical latitudes today, as was S. catatumbus during the Cenozoic. Previous studies suggest that the biogeography of Fabaceae is related to temperature, with higher abundances in warmer environments. We plan to assess the relationship between abundance and temperature in both the modern and fossil taxa.

Our study provides a snapshot of the diversification of Fabaceae, the most abundant plant family in the Neotropics, and demonstrates that the use of Airyscan microscopy and machine learning allows us to extract phylogenetic information from pollen morphology that had been unattainable previously. Given estimates that >50% of fossil palynomorphs have unknown taxonomic affinities or are only known to belong to an order or family, our results present a powerful new approach to analyzing palynological data.